A Paris nonprofit called Current AI wants to do for AI what the early World Wide Web did for documents: make a shared public layer that anyone can use, without paying rent to a handful of companies. It launched in February 2025 at the AI Action Summit with a $400 million starting endowment and a longer target of $2.5 billion over five years, and it has been quietly shipping the first pieces of that stack ever since.
The pitch matters more than the funding number. If you can already run a competent open-weight model on a laptop, what is missing is not the model. It is the public data, the shared compute, the accountability tools, and the local-language coverage that big vendors have little reason to build. TechCrunch profiled the effort on July 19, and the gap is the reason the nonprofit keeps talking about the early web rather than about a model release.
The pitch is the early Web, not a model launch
Current AI was founded by Martin Tisné, who also chairs the Omidyar Network-funded AI Collaborative. Fortune reported at the Paris launch that the initial $400 million endowment was assembled from France, several other governments, Google, Salesforce, and a group of large US foundations. The longer $2.5 billion target is what the organization says it wants to mobilize by 2030. A follow-up TechCrunch profile on July 19 said France had seeded the nonprofit with $100 million.
The Paris Charter on AI in the Public Interest, signed at the same summit, makes the framing concrete. It commits signatories to AI systems that are “open, diverse, sustainable, locally relevant, development-centered and accessible around the world,” and lists three core principles: openness, accountability, and participation and transparency. The countries that signed on February 11, 2025 were Chile, Finland, France, Germany, India, Kenya, Morocco, Nigeria, Slovenia, and Switzerland. The launch drew public remarks from United Nations Secretary-General Antonio Guterres, European Commission President Ursula von der Leyen, and French President Emmanuel Macron, who endorsed Current AI’s public-interest mission. According to the TechCrunch profile, the French government seeded the nonprofit with $100 million of that total.
Ayah Bdeir, who joined as CEO in January 2026 after running Mozilla’s AI strategy as a senior advisor, told TechCrunch the framing in one line: “If AI is truly a transformative technology…there has to be a public alternative.” The same piece quotes her describing the goal as “Like the World Wide Web, available to anyone, for free.” Bdeir’s track record makes the pitch concrete: she founded the electronics kit company littleBits in 2011, raised more than $47 million in venture funding, and sold the company to Sphero in August 2019 before moving into public-interest technology.
Three work streams, not one big model
Current AI organizes the work as three streams that the organization’s site labels Fund, Build, and Bridge in its navigation. Fund is grant money to local groups. Build is the open infrastructure, from hardware to applications. Bridge is shared collaboration tooling. The organization’s next-chapter blog post frames the start-up phase around three priorities: supporting openness, unlocking data, and ensuring accountability. The point is that no single stream is the product; the public-interest alternative is the whole stack taken together.
The grants are the part with the most deployed code so far. TechCrunch reports Current AI has put $3.2 million into four organizations across Kenya, Lebanon, and the Brazilian Amazon: Masakhane (AI datasets for more than 50 African languages), the Institute for Worldmaking (digitizing Arab cultural history in Lebanon), Portal sem Porteiras (offline AI tools with Indigenous Amazon communities), and the African Internet Rights Alliance (AI accountability audit tools). Bdeir described the intended shape of this work in the same interview: “This could look like an Indigenous elder in the Brazilian Amazon using a tool built in Kenya to be able to pass down ecological knowledge in their own language.”
There are also two flagship hardware and software projects. Suno Sutra is an offline device that runs AI in 22 Indian languages, built with India’s Bhashini program and shown at the India AI Summit in February 2026. Alpha Chat is an open-source chatbot that Bdeir’s team says was built in seven weeks by ten organizations including Hugging Face, Mozilla, and the MIT Media Lab, and launched at the AI for Good Summit in Geneva in early July 2026. A separate partnership with Tokyo-based Sakana AI targets shared open-source infrastructure for Japanese and other Global South language communities. The organization’s Gap Map v0.1, published on July 1, 2026, is a working inventory of the existing open-source AI stack that the project is meant to fill in.
What is missing from the stack today
The honest gap is data, not models. Bdeir put it bluntly in the TechCrunch interview: “Big tech builds multilingual models to expand their market,” she said, “regardless of consent or context.” She added, “And with English driving the largest language models and AI systems, a bulk of the world’s languages…are left behind.” The Current AI thesis is that public-interest languages and public-interest datasets have to be built deliberately, by groups that the host companies have no reason to fund.
The funders also frame themselves differently from investors. “They’re not investors; they’re funders,” Bdeir told TechCrunch. The same piece captures her rejecting the size-first worldview directly: “Scale is not always the measure. That is the Big Tech model.” The structure that follows from that is grant money for local groups, shared open infrastructure rather than a flagship model, and accountability tools so that governments and nonprofits can audit what they are running. The Current AI next-chapter post lists openness, data access, and accountability as the three priorities of what it calls its “start-up phase.”
The funding roster reinforces that framing. Alongside the $100 million euros from France and the $400 million endowment, the named backers include the Ford Foundation, MacArthur Foundation, and McGovern Foundation on the philanthropy side, with Google and Salesforce as the lead tech partners. Hugging Face, Instacart, and Sakana AI appear as supporters in the Fortune launch report. DeepMind is named as a backer in the TechCrunch follow-up.
What This Means
The early-web analogy does a lot of work for Current AI. The original web solved three problems at once: a shared protocol, public infrastructure that anyone could host on, and a permission to link without asking. Current AI is trying to assemble the same three layers for AI: open weights and open data as the protocol, public-interest compute and grant-funded local groups as the infrastructure, and accountability tools and audit frameworks as the linking layer. The first two are largely in motion, and the Gap Map v0.1 is the project’s own attempt to publish what’s missing.
For local-AI readers, the practical question is whether the funded work arrives as something you can actually run. Masakhane’s datasets and Alpha Chat’s chatbot are the most usable of the named projects so far. Suno Sutra is a hardware device, not a download. The gap between the $400 million endowment and a working “World Wide Web of AI” remains real, but the gap map and the seven-week chatbot build suggest the team is willing to ship small pieces rather than wait for a flagship.
The Bottom Line
Current AI is a $400 million bet that the missing layer in AI is public-interest infrastructure rather than another frontier model, and that the missing languages matter as much as the missing benchmarks. The Paris charter, the named grants, and the open-source chatbot are the proof so far; the rest is a five-year build.